To explore this question, I tried to build a data set of the testing history of several comparable countries and then put them on a common timeline with the US; This is easier said than done for several reasons.
- Some major countries like Spain, China, and Germany don't appear to publish their testing data at all.
- Most countries only update a page with that day's count making it very difficult to know how many tests they've done in past days
- When countries provide a number of tests, its often unclear whether the number represents samples tested or unique people tested. This is important because many tests require two samples per person so a country that reports 20,000 tests may only have tested 10,000 people.
- The US has split their test counts between the ones done at the CDC and the one done by private labs like Labcorp and Quest that are counted separately.
I was able to get testing data from most countries I'm tracking. I added the testing data fields to the Covid Raw Data Spreadsheet. All my sources can be viewed by mousing ovedr the orange triangles in the data labels on the left side of the page (if you know of better sources or see a mistake, please comment in this blog). For the US testing info, I was originally relying on the CDC's Covid-19 Testing Page, but then I found my way to the Covid Tracking Project , where a stout team of volunteers check in with all US states and territories to get daily results that they post on their useful site. Their numbers are lower than CDCs because they count persons while the CDC counts tests, but their data has been used by the NYT and other major news outlets and seems very trustworthy. HT also to Our World In Data's test source page which pointed me to many country sources for testing reporting data.
The Covid Raw Data sheet has begun tracking testing counts as of today and will be updated every morning, but there were only a few countries where I could recreate testing history by scanning past press releases and bulletins from a few countries nice enough to archive them. Should anyone discover another source for prior testing count data for other countries, please let me know.
With that out of the way, let's look at the data. The first thing I looked at was what our current testing situation is relative to other countries. Here's the current test count info from the testing data tab:
Note absolute test and new test info. Note that new test info is promising, but what did we miss?
{Note: Graph updated to reflect new info}
Note this is just the testing counts. Hopefully we won't match their case counts! South Korea is way ahead of us in testing, which isn't surprising, but we've matched their daily pace. As we'd expect, many other countries have testing counts way ahead of where we were a week ago around day 9, but how are we doing on pace? Let's look at the next graph where we look at the number of new tests daily:
This chart shows how every country can vary pretty wildly on a day to day basis with their testing, but in the last couple days the US has done more tests in a day than any country in my tables.
Now that we have a sense of the current situation, you're probably wondering if the time we lagged on testing means there are many cases in the US that we're blind to. Perhaps we're much worse off than we know. That's a good question. Let's see what we can figure out. In my last post, I used fatality reports as a way to check on testing data. If the US were under-testing compared to other countries, you would still expect them to correctly report fatalities from Covid. So let's look at the growth of fatalities on the common timeline:
Some of the lines are broken where I don't have data in my table, but the US seems squarely in the middle of the other country's reports. This suggests the US case counts aren't likely to be very distorted by the early slow in testing compared to other counties.
So to sum up, we appear to have gained significant ground on testing and are now testing at levels as high or higher than any other country. Also the lapse in early tests seems unlikely to be hiding a mass number of cases that other countries detected on their timelines.
That said, its clear we should have done better at being more prepared for this epidemic and at changing our testing protocols to adjust to the circumstances. Also, while we don't seem to be behind other countries on testing at this point, that doesn't mean we have an accurate picture of our situation. It's quite likely that there are many times more people with COVID-19 than testing shows. We'll look at that in a future posting, but at the moment, I want to show where we are relative to other countries.
Have a question? Disagree? Drop a comment.
{Note: Graph updated to reflect new info}
Note this is just the testing counts. Hopefully we won't match their case counts! South Korea is way ahead of us in testing, which isn't surprising, but we've matched their daily pace. As we'd expect, many other countries have testing counts way ahead of where we were a week ago around day 9, but how are we doing on pace? Let's look at the next graph where we look at the number of new tests daily:
This chart shows how every country can vary pretty wildly on a day to day basis with their testing, but in the last couple days the US has done more tests in a day than any country in my tables.
Now that we have a sense of the current situation, you're probably wondering if the time we lagged on testing means there are many cases in the US that we're blind to. Perhaps we're much worse off than we know. That's a good question. Let's see what we can figure out. In my last post, I used fatality reports as a way to check on testing data. If the US were under-testing compared to other countries, you would still expect them to correctly report fatalities from Covid. So let's look at the growth of fatalities on the common timeline:
Some of the lines are broken where I don't have data in my table, but the US seems squarely in the middle of the other country's reports. This suggests the US case counts aren't likely to be very distorted by the early slow in testing compared to other counties.
So to sum up, we appear to have gained significant ground on testing and are now testing at levels as high or higher than any other country. Also the lapse in early tests seems unlikely to be hiding a mass number of cases that other countries detected on their timelines.
That said, its clear we should have done better at being more prepared for this epidemic and at changing our testing protocols to adjust to the circumstances. Also, while we don't seem to be behind other countries on testing at this point, that doesn't mean we have an accurate picture of our situation. It's quite likely that there are many times more people with COVID-19 than testing shows. We'll look at that in a future posting, but at the moment, I want to show where we are relative to other countries.
Have a question? Disagree? Drop a comment.
Hey Chris -
ReplyDeleteReally love what you're doing here. One thing that could be introducing noise into your fatalities on a common timeline analysis is the relative age of case clusters and/or the population overall for each country.
For example, a death among the 65+ population would suggest ~5.5 additional cases (assuming a 15% mortality rate)
However a death from someone in the 20-29 age range would suggest ~500 additional cases (assuming a 0.2% mortality rate).
Since cases are still (relatively) rare, one case cluster in a nursing home can really skew the fatality curve up if we aren't adjusting for the age factor. This should smooth out as case exposure grows and assumes a more normal distribution.
I don't know if there is a way to get access to average age of case data, but that could help refine the conclusions drawn from the total fatality numbers.
Best,
Josh